Enable optimized dynamic quantization on aarch64 #126687
Open
+13
−4
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
oneDNN+ACL has optimized kernels for s8s8 matmul, so input is signed. This change leaves behaviour on all other platforms the same. This change requires intel/ideep#313 to go in, and oneDNN 3.5 for the optimized kernels. This change speeds up dynamic quantized linear by ~10x.
jondea
requested review from
jerryzh168,
salilsdesai,
kimishpatel,
digantdesai and
jianyuh
as code owners
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/126687
Note: Links to docs will display an error until the docs builds have been completed. ✅ No FailuresAs of commit 9469bae with merge base afda668 ( This comment was automatically generated by Dr. CI and updates every 15 minutes. |
|
|
pytorch-bot
bot
added
module: cpu
drisspg
added
the
triaged
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
oneDNN+ACL has optimized kernels for s8s8 matmul, so input is signed. This change leaves behaviour on all other platforms the same. This change requires intel/ideep#313 to go in, and oneDNN 3.5 for the optimized kernels. This change speeds up dynamic quantized linear by ~10x.
Also, do you have a policy on copyright headers? Arm's usual policy when contributing to open source projects is to include a copyright header on any file which is modified. Would this be acceptable? If not, is there somewhere else suitable to note copyright?
cc @jgong5 @mingfeima @XiaobingSuper @sanchitintel @ashokei @jingxu10